In this paper, we introduce FairSense-AI: a multimodal framework designed to detect and mitigate bias in both text and images. By leveraging Large Language Models (LLMs) and Vision-Language Models (VLMs), FairSense-AI uncovers subtle forms of prejudice or stereotyping that can appear in content, providing users with bias scores, explanatory highlights, and automated recommendations for fairness enhancements. In addition, FairSense-AI integrates an AI risk assessment component that aligns with frameworks like the MIT AI Risk Repository and NIST AI Risk Management Framework, enabling structured identification of ethical and safety concerns. The platform is optimized for energy efficiency via techniques such as model pruning and mixed-precision computation, thereby reducing its environmental footprint. Through a series of case studies and applications, we demonstrate how FairSense-AI promotes responsible AI use by addressing both the social dimension of fairness and the pressing need for sustainability in large-scale AI deployments. https://vectorinstitute.github.io/FairSense-AI, https://pypi.org/project/fair-sense-ai/ (Sustainability , Responsible AI , Large Language Models , Vision Language Models , Ethical AI , Green AI)

In the age of increasingly realistic generative AI, robust deepfake detection is essential for mitigating fraud and disinformation. While many deepfake detectors report high accuracy on academic datasets, we show that these academic benchmarks are out of date and not representative of real-world deepfakes. We introduce Deepfake-Eval-2024, a new deepfake detection benchmark consisting of in-the-wild deepfakes collected from social media and deepfake detection platform users in 2024. Deepfake-Eval-2024 consists of 45 hours of videos, 56.5 hours of audio, and 1,975 images, encompassing the latest manipulation technologies. The benchmark contains diverse media content from 88 different websites in 52 different languages. We find that the performance of open-source state-of-the-art deepfake detection models drops precipitously when evaluated on Deepfake-Eval-2024, with AUC decreasing by 50\% for video, 48\% for audio, and 45\% for image models compared to previous benchmarks. We also evaluate commercial deepfake detection models and models finetuned on Deepfake-Eval-2024, and find that they have superior performance to off-the-shelf open-source models, but do not yet reach the accuracy of deepfake forensic analysts. The dataset is available at https://github.com/nuriachandra/Deepfake-Eval-2024.

The ever-growing availability of unlabeled data presents both opportunities and challenges for training artificial intelligence systems. While self-supervised learning (SSL) has emerged as a powerful paradigm for extracting meaningful representations from vast amounts of unlabeled data, existing methods still struggle to adapt to the non-stationary, non-IID nature of real-world data streams without forgetting previously learned knowledge. Recent works have adopted a repeated cosine annealing schedule for large-scale continual pre-training; however, these schedules (1) inherently cause forgetting during the re-warming phase and (2) have not been systematically compared to existing continual SSL methods. In this work, we systematically compare the widely used cosine schedule with the recently proposed infinite learning rate schedule and empirically find the latter to be a more effective alternative. Our extensive empirical evaluation across diverse image and language datasets demonstrates that the infinite learning rate schedule consistently enhances continual pre-training performance compared to a repeated cosine decay without being restricted to a fixed iteration budget. For instance, in a small-scale MAE pre-training setup, it outperforms several strong baselines from the literature. Our results show that the infinite learning rate schedule remains effective at scale, surpassing repeated cosine decay for both MAE pre-training and zero-shot LM benchmarks. These models are known for their massive parameter counts and extensive training on vast amounts of data, often developing impressive general-purpose capabilities unexpectedly during pre-training (Brown et al., 2020; Wei et al., 2022). While foundation models have demonstrated remarkable success on static tasks, adapting them to evolving data--such as the continuous influx of new textual information (Soldaini et al., 2024; Li et al., 2024; Abadji et al., 2022; Kocetkov et al., 2022) and the emergence of novel visual concepts (Prabhu et al., 2023; Seo et al., 2024)--remains a major challenge.

The in-context learning capabilities of modern language models have motivated a deeper mathematical understanding of sequence models. A line of recent work has shown that linear attention models can emulate projected gradient descent iterations to implicitly learn the task vector from the data provided in the context window. In this work, we consider a novel setting where the global task distribution can be partitioned into a union of conditional task distributions. We then examine the use of task-specific prompts and prediction heads for learning the prior information associated with the conditional task distribution using a one-layer attention model. Our results on loss landscape show that task-specific prompts facilitate a covariance-mean decoupling where prompt-tuning explains the conditional mean of the distribution whereas the variance is learned/explained through in-context learning. Incorporating task-specific head further aids this process by entirely decoupling estimation of mean and variance components. This covariance-mean perspective similarly explains how jointly training prompt and attention weights can provably help over fine-tuning after pretraining.

Deep neural networks (DNNs) have brought significant advancements in various applications in recent years, such as image recognition, speech recognition, and natural language processing. In particular, Vision Transformers (ViTs) have emerged as a powerful class of models in the field of deep learning for image classification. In this work, we propose a novel Random Matrix Theory (RMT)-based method for pruning pre-trained DNNs, based on the sparsification of weights and singular vectors, and apply it to ViTs. RMT provides a robust framework to analyze the statistical properties of large matrices, which has been shown to be crucial for understanding and optimizing the performance of DNNs. We demonstrate that our RMT-based pruning can be used to reduce the number of parameters of ViT models (trained on ImageNet) by 30-50\% with less than 1\% loss in accuracy. To our knowledge, this represents the state-of-the-art in pruning for these ViT models. Furthermore, we provide a rigorous mathematical underpinning of the above numerical studies, namely we proved a theorem for fully connected DNNs, and other more general DNN structures, describing how the randomness in the weight matrices of a DNN decreases as the weights approach a local or global minimum (during training). We verify this theorem through numerical experiments on fully connected DNNs, providing empirical support for our theoretical findings. Moreover, we prove a theorem that describes how DNN loss decreases as we remove randomness in the weight layers, and show a monotone dependence of the decrease in loss with the amount of randomness that we remove. Our results also provide significant RMT-based insights into the role of regularization during training and pruning.

The rapid growth of Large Language Models (LLMs) presents significant privacy, security, and ethical concerns. While much research has proposed methods for defending LLM systems against misuse by malicious actors, researchers have recently complemented these efforts with an offensive approach that involves red teaming, i.e., proactively attacking LLMs with the purpose of identifying their vulnerabilities. This paper provides a concise and practical overview of the LLM red teaming literature, structured so as to describe a multi-component system end-to-end. To motivate red teaming we survey the initial safety needs of some high-profile LLMs, and then dive into the different components of a red teaming system as well as software packages for implementing them. We cover various attack methods, strategies for attack-success evaluation, metrics for assessing experiment outcomes, as well as a host of other considerations. Our survey will be useful for any reader who wants to rapidly obtain a grasp of the major red teaming concepts for their own use in practical applications.

Fox News senior national correspondent Aishah Hasnie breaks down Democrats' coordinated protests against President Donald Trump's speech to Congress on'Special Report.' Conservatives on social media slammed Senate Democrats for posting videos with identical scripts ahead of President Donald Trump's joint address to Congress Tuesday night. Mashups of the identical videos, which included Sen. Cory Booker, D-N.J., leading the "Sh-- That Ain't True" social media campaign, have gone viral on social media. Many conservatives on social media, including Elon Musk, are asking, "Who is writing the words that the puppets speak? Senators Elizabeth Warren, D-Mass., and Chuck Schumer, D-N.Y., led the charge with their matching social media videos this morning. About two dozen Senate Democrats have since followed with their own identical posts. The video begins with a clip of Trump vowing to "bring prices down starting on day one" followed by a cut-in of the Senate Democrats saying: "Sh-- That Ain't True?

Last week, standing in front of President Donald Trump's first Cabinet meeting, Elon Musk, the unelected billionaire running a blitzkrieg against the US government, acknowledged that he had made a mistake--that in going after the US Agency for International Development, the foreign assistance program that he has all but destroyed, he accidentally ended the Ebola prevention project it ran overseas. Musk claimed the error was quickly fixed and there was no interruption in service. But former and current USAID staff quickly told the Washington Post that Musk was wrong--the Ebola response remained sharply curtailed. And, as the Bulwark reported, Nicholas Enrich, the acting assistant administrator for global health at USAID, who was placed on administrative leave Sunday, had drafted an unfinished memo that predicted the demolition of USAID would lead to more than 28,000 cases of Ebola and related diseases, as well as a 28 to 32 percent increase in tuberculosis globally, up to 18 million cases of malaria (with up to 166,000 deaths annually), and an additional 200,000 cases of paralytic polio a year. Musk's assertion that his slash-and-burn assault on USAID had no negative impact on combating Ebola was disinformation. He was hiding the truth on a critical global health issue.

The filing in question was related to a case in which Guyer's client Karen Iovino claimed she faced retaliation from employer Michael Stapleton Associates, and "was fired for reporting alleged issues about MSA's contract with the State Department to that agency's Office of Inspector General." In an August filing, Guyer denied citing "'fictitious' cases," and said that the cases did in fact exist, but that they were misquoted and miscited by generative AI. "GPTs generate excellent to brilliant legal arguments," Guyer wrote in a separate declaration, saying that the errors were generated by Atrophic Inc.'s Claude 3 Opus, which is one of several AI tools Guyer says he uses. "I utilize a suite of generative AI technologies for legal research and writing purposes, and GPT legal document briefing," Guyer said.

European Commission (EC) President Ursula von der Leyen has proposed a five-part plan to mobilise some 800bn euros ( 842bn) to beef up Europe's defence and provide "immediate" military support to Ukraine after the United States suspended aid. "A new era is upon us," the president said in a letter presenting the plan to 27 European Union (EU) leaders on Tuesday, two days before a summit aimed at cementing joint action on Ukraine and Europe's long-term security begins in Brussels. "Europe faces a clear and present danger on a scale that none of us has seen in our adult lifetime," she wrote. European leaders are under huge pressure to increase defence spending as US President Donald Trump's return to power has delivered a rude wake-up call that they cannot blindly rely on Washington. The joint borrowing would go towards building pan-European capability domains like air and missile defence, artillery systems, missiles and ammunition, drones and anti-drone systems or to address other needs from cyber- to military mobility, the EC said.